Virtual intimacy consumes physical resources

Believe virtual intimacy is an eco-conscious alternative to traditional dating? Think twice. Regrettably, the servers supporting our bot partners require substantial energy and water. Essentially, the AI landscape faces a significant sustainability challenge, especially with the rising adoption of chatbot companions globally.

The significant water usage linked to AI is clarified in a FluixAI piece. It’s attributed to the evaporative HVAC systems crucial in data centers. These systems rely on water evaporation to cool air, subsequently cooling servers. As air interacts with water-saturated pads, heat is absorbed and water evaporates, consuming a substantial amount of water.

In essence, this underscores the tangible resources tied to our intangible digital engagements.

Alarming figures: AI’s impact goes beyond mere water consumption

According to a study led by Pengfei Li and fellow researchers from UC Riverside and UT Arlington, training GPT-3 in Microsoft’s advanced U.S. data centers can directly evaporate 700,000 liters of fresh water. The study highlights the secrecy surrounding such information. It reveals that «warehouse-scale data centers» consume vast amounts of water, with Google’s self-owned centers alone withdrawing 25 billion liters and using nearly 20 billion liters for on-site cooling in 2022, most of which was potable water.

Another staggering statistic shows that Google, Microsoft, and Meta withdrew 2.2 billion cubic meters of water in 2022, equivalent to the total annual water withdrawal of two Denmark-sized entities. The U.S. accounted for 1.5 billion cubic meters of this withdrawal, estimated at 0.33% of the total U.S. annual water withdrawal. The study projects that by 2027, AI will globally withdraw more water than «4-6 Denmark-sized entities or half of the United Kingdom,» approximately 4.2-6.6 billion cubic meters.

Potential remedies, but little tangible progress as of now

A previous paper from 2023 examines AI’s direct and indirect water usage, encompassing cooling and microchip production, as well as the water footprint of electricity used in data center operations. The authors advocate for enhancing energy efficiency, adopting renewable energy sources, optimizing algorithms, and instituting water conservation strategies. However, they acknowledge substantial hurdles in terms of research and implementation.

Despite rapid advancements in AI during the early months of 2024, resulting in increased water consumption due to the introduction of numerous models for industrial and consumer applications, there’s a positive shift toward addressing sustainability concerns within the AI sector. This heightened awareness may prompt action from policymakers and environmental organizations.

The consequences of AI-based digital intimacy remain unaccounted for

There’s a glaring omission in the discussion, one that even the industry giants seem to overlook—the substantial impact on water usage stemming from the increasing number of individuals engaging extensively with their chatbot partners.

Such usage patterns remain unaddressed in reports like Stanford University’s recent AI Index Report, an annual publication from the Human Centered Artificial Intelligence department.

FluixAI highlighted a concerning statistic: «for every 20 to 50 questions answered, ChatGPT’s servers indirectly consume the equivalent of a 16.9 oz water bottle.»

While this figure lacks specifics regarding the length of ChatGPT or other AI responses, or the methodology used to estimate consumption, it underscores the hidden water expenditure associated with AI-mediated intimacy.

What’s needed is data. A study focusing on digisexual behavior could provide insights into users’ engagement duration with intimate chatbot companions, the average duration of digital encounters, and their frequency per week (or day). With estimates of user numbers per chatbot company, we could begin to grasp the water-related costs of AI-mediated intimacy.

I am text block. Click edit button to change this text. Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.

What actions can we take?

According to the 2024 AI Index Report, AI has surpassed human capabilities in various benchmarks, such as image classification, visual reasoning, and English comprehension, yet it still lacks human-style «commonsense planning and reasoning.»

I suggest that humans could leverage our innate ability for commonsense planning by openly addressing and considering the impact of human-AI sexual interactions. This isn’t about adopting a sex-negative stance, but rather about acknowledging a significant aspect of AI usage as responsible adults. By doing so, we can collectively address sustainability concerns without stigma, fostering collaboration among all users, including those partnered with AI companions.

As FluixAI emphasizes, «Every interaction with AI contributes to a broader narrative that encompasses not only data and algorithms but also water, energy, and the environment. As consumers, we can influence change by supporting companies committed to sustainability.»

Ultimately, it’s a choice between proactive engagement in solving sustainability challenges or potentially facing the dilemma of telling our AI partners, «not tonight, dear, I have a headache… from water scarcity.»